1. Field of the Invention
The present invention generally relates to a questionless case-based knowledge base accessible by an intelligent search engine and a method for constructing the same from pre-existing on-line documentation.
2. Description of Related Art
Expert systems are comprised of two parts, a knowledge base and an engine. Traditionally, the engine has either been a reasoning or inference engine which embodies a problem-solving method or procedure and uses the knowledge in the knowledge base to construct a line of reasoning which leads to a solution for the problem. The most common line of reasoning used by an expert system involves the chaining, either forward, backward or a flexible mix thereof, of IF-THEN rules. However, as knowledge of the domain for a particular problem is almost always incomplete and, has, therefore, a degree of uncertainty in the solution thereof, a rule may have associated therewith, a confidence factor ("CF") or weight. Alternately, using "fuzzy logic", the degree of uncertainty associated with a rule may be represented by a distribution of values. Using the CFs or uncertainty distribution, the inference engine is able to evaluate various lines of reasoning and provide probabilities of correctness for the various lines of reasoning.
Typically, the knowledge base of an expert system is organized in a specific representational form for use by the inference engine. One such system, generally referred to as a rule-based system, arranges knowledge as a series of rules, each consisting of an IF part and a THEN part. The IF part lists a set of conditions in some logical combination. The piece of knowledge represented by the rule is relevant to the line of reasoning being developed if the IF part of the rule is satisfied. Consequently, the THEN part can then be concluded, or its action taken.
A related representational form of a knowledge base which is more suitable for use in complex systems is generally referred to as a "case-based" knowledge base. In this format, knowledge is arranged as a series of discrete record entities commonly known as cases. Generally, a knowledge base would be provided with a case for each problem to be addressed by the associated expert system. Each case is structured to include a title, a description field, a list of questions and answers, and a solution. Various case-based systems which are similar to the system described above and which are presently commercially available include CBR Express and CasePoint, both of which are manufactured by Inference Corporation of El Segundo, Calif.
When executing an operation, for example, determining a solution to a problem, using a case-based reasoning system, the user of the system is required to type in a natural language description of a symptom of the problem. Using the symptom provided by the user, the engine scans all of the questions residing in the knowledge base and returns with a list of questions to be answered by the user. Based upon the answers to the proffered questions, the engine narrows the search to a solution set forth in one of the cases.
Expert systems which utilize case-based knowledge bases have several shortcomings, particularly with respect to the question/answer list. In order to function properly, the question/answer list must be carefully crafted such that, based upon the answers to the questions, the engine is led to the correct solution. As a result, formation of the question/answer portion of a case-based knowledge base is a critical element of the case building process which is very manpower intensive and often requires on the order of 70-80% of the total time required to build the knowledge base for an expert system. Due to the demands required to build such a knowledge base, there is often insufficient manpower to fully test and fine-tune the expert system.
Another problem with, expert systems which incorporate a case-based knowledge base is that such systems have been traditionally used in the so-called "help-desk" environment where the actual user of the device would verbally describe the symptoms over the phone to a trained technician at a remote location. The technician would then relay the questions generated by the expert system to the user. As the technician was much more familiar with both the expert system and the knowledge base, unclear or difficult questions could be explained to users with relatively little experience with the expert system. If, however, the expert system were installed in the user's computer system, it is entirely possible that the user would be unable to fully understand the questions. Such questions, if improperly answered, could potentially impede the expert system from being able to properly diagnose a problem.
While it would be desirable to make such automated expert systems broadly available, the cost and "ease of use" problem presented by the expert system's use of the question/answer list remain serious impediments to acceptance of such systems by the general public. Due to these shortcomings, computer users have traditionally relied upon either the "help-desk" or upon operating manuals published by the manufacturer for assistance when faced with an operating problem.
With the advent of CD-ROM, computer users have become increasingly comfortable with the concept of reviewing documents on-line. Accordingly, on-line documentation has become increasingly available in recent years and has been lost popular with those computer users who prefer avoiding having to continuously reference printed publications. In the near future, it is anticipated that many computer users will expect that on-line documentation accompany all of their hardware or software purchases. Furthermore, as the additional costs involved in making printed materials available on-line are minimal, the computer industry will likely satisfy this demand without hesitation.
With the increasing availability of on-line documentation, it can be readily seen that it would be desirable to be able to modify such documentation to provide a questionless case-based knowledge base suitable for use with a system which incorporates an intelligent search engine and which will provide the user with a solution to a problem without requiring the user to correctly answer a series of questions generated by the system. Accordingly, it is an object of this invention to provide such a technique.